Transformer models can be used to process a sequence-to-sequence information and effectively capture dependencies among words or tokens in the sequence. ECG signals are the time sequence data. The transformer model is well suited for classification of ECG signals. In this paper, we use the encoder architecture of the transformer model to detect heart rate variability and classify electrocardiogram (ECG) data from the PhysioNet. The results show that our model can achieve 91% accuracy, 94% recall, 92% precision and 93% F1 Score. The experimental results demonstrate that the proposed method achieves a good classification performance.